Belt-Tracking Control: A Linear Quadratic Gaussian Approach

An endless-web active tracking system has been designed, built, and tested on an experimental imaging apparatus. A continuous steady-state Kalman filter is used to estimate the 3rd-order system states. From noisy measurement of belt edge position only (no roller tilt measurement), estimates of all three state variables are generated by the Kalman filter. All three state estimates are utilized in the optimal control applied to the steering roller servomotor. The optimal control is designed to minimize a quadratic performance index which penalizes deviations of the belt edge position and steering roller tilt. The new tracking system compares favorably with the previous system in performance tests and does not require the potentiometer for roller tilt measurement essential in the previous system.